Compositional pattern producing networks: A novel abstraction of development
read more
Citations
Principles of Neural Science
Deep neural networks are easily fooled: High confidence predictions for unrecognizable images
NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis
Regularized Evolution for Image Classifier Architecture Search
Large-scale evolution of image classifiers
References
Genetic Programming: On the Programming of Computers by Means of Natural Selection
Approximation by superpositions of a sigmoidal function
The Chemical Basis of Morphogenesis
Principles of Neural Science
Principles of Neural Science
Related Papers (5)
A hypercube-based encoding for evolving large-scale neural networks
Frequently Asked Questions (6)
Q2. What are the future works in "Compositional pattern producing networks: a novel abstraction of development" ?
To extend this notion of environmental embodiment further only requires that other position-dependent environmental features are input to the CPPN as well. Just as methods for mapping other developmental encodings to useful phenotypes have motivated much research to date, methods for CPPN interpretation are fruitful avenues for future research as well. That is, the CPPN can be re-queried at every point ( or at every point where change has occurred ). This capability supports further that CPPNs belong in the same class as other developmental encodings.
Q3. What is the main assumption behind the work in developmental encoding?
A strong assumption behind much recent work in developmental encoding is that local interaction and temporal unfolding in a simulation are abstractions that capture the essential properties of development in nature.
Q4. Why was it necessary to establish that CPPNs belong in the same class of abstractions?
Because CPPNs are unlike traditional developmental encodings, it was necessary to establish that they nevertheless belong in the same class of abstractions.
Q5. Why is it important to simulate every facet of a biological process?
because computers differ from the natural world and because it is not necessary to simulate every facet of a biological process in order to capture its essential properties, finding the right level of abstraction remains an open problem.
Q6. How can the authors explain the evolution of coordinate frames without a notion of time?
Through function composition, it is simple to demonstrate that biologically plausible new coordinate frames are derivable from preexisting ones without a notion of time.